Your First MCP Server Monitor

By the end of this guide, you’ll have an MCP server monitor verifying that your AI tool server starts, connects over the Model Context Protocol, and exposes the expected tools.

Prerequisites

DevHelm CLI installed or an API token for REST calls
An API token set as DEVHELM_API_TOKEN — see Authentication
An MCP server package (e.g., published to npm) or a command that starts one

Create the monitor

Create the MCP server monitor

devhelm monitors create \
  --name "My MCP Server" \
  --type MCP_SERVER \
  --url npx \
  --frequency 300 \
  --regions us-east

version: "1"
monitors:
  - name: My MCP Server
    type: MCP_SERVER
    config:
      command: npx
      args:
        - "@my-org/mcp-server"
      env:
        API_KEY: ${MCP_API_KEY}
    frequencySeconds: 300
    regions:
      - us-east

curl -X POST https://api.devhelm.io/api/v1/monitors \
  -H "Authorization: Bearer $DEVHELM_API_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{
    "name": "My MCP Server",
    "type": "MCP_SERVER",
    "config": {
      "command": "npx",
      "args": ["@my-org/mcp-server"],
      "env": {
        "API_KEY": "your-api-key"
      }
    },
    "frequencySeconds": 300,
    "regions": ["us-east"]
  }'

With the CLI, --url is the command that launches the MCP server. Arguments and environment variables aren’t settable via flags — use YAML config-as-code (config.args, config.env) or the API for the full configuration, as shown in the other tabs.

Verify it's running

devhelm monitors list --type MCP_SERVER
devhelm monitors get <monitor-id>

Check results

Results show whether the MCP server started, connected, and listed its capabilities.

devhelm monitors checks <monitor-id> --limit 5

Add assertions

Verify that the MCP server exposes the tools and resources you expect:

monitors:
  - name: My MCP Server
    type: MCP_SERVER
    config:
      command: npx
      args:
        - "@my-org/mcp-server"
    frequencySeconds: 300
    regions:
      - us-east
    assertions:
      - config:
          type: mcp_connects
        severity: fail
      - config:
          type: mcp_tool_available
          toolName: search
        severity: fail
      - config:
          type: mcp_min_tools
          min: 3
        severity: warn

Key MCP assertions:

Type	What it checks
`mcp_connects`	Server starts and completes the MCP handshake
`mcp_tool_available`	A specific tool is listed in the server’s capabilities
`mcp_tool_count`	Server exposes at least N tools
`mcp_resource_available`	A specific resource is available
`mcp_startup_time`	Server starts within a time threshold

For the full list, see MCP Server configuration.

How it works

Each check cycle:

DevHelm starts the MCP server using the configured command and args
Connects via the Model Context Protocol (stdio transport)
Requests the server’s tool and resource listings
Evaluates assertions against the response
Terminates the server process

Environment variables in env are passed to the server process. Use DevHelm secrets (${VAR} syntax) to avoid storing credentials in config files.

Next steps

MCP Server reference

Full configuration and assertion details.

First alert

Get notified when your MCP server fails.

MCP Server SDK

Use DevHelm as an MCP server in AI agents.

Overview

Getting Started

Monitoring Guides

Alerting Guides

Incident Guides

Status Data Guides

Monitoring as Code Guides

Your First MCP Server Monitor

Create the monitor

Add assertions

How it works

Next steps

MCP Server reference

First alert

MCP Server SDK

​Create the monitor

​Add assertions

​How it works

​Next steps

MCP Server reference

First alert

MCP Server SDK

Create the monitor

Add assertions

How it works

Next steps